skills: add pre-main-push gate

[VinciGit00]

Summary

• Adds .claude/skills/pre-main-push.md, a Claude Code skill that blocks direct pushes to main and enforces: (1) user-supplied API key per session, (2) 5-URL validation of the chosen v2 endpoint, (3) PR-only workflow.
• Encodes the v1→v2 endpoint mapping (markdownify→scrape, smartscraper→extract, searchscraper→search, smartcrawler→crawl.start) and pins the v2 host v2-api.scrapegraphai.com/api/*.
• Hard-stops if the key is missing, any URL fails, or gh isn't authenticated. Forbids logging the key or bypassing with --force/--no-verify.

Validation

Endpoint extract tested against 5 live websites via tests/python-v2.1.0/test_extract.py — 5/5 returned status=success (example.com, scrapegraphai.com, iana.org, example.org, httpbin.org/html). API key supplied by user at PR creation time; not stored or logged.

Test plan

• Skill triggers on "push to main" / "merge to main" / /pre-main-push
• Key is requested each invocation; never read from .env
• 5/5 endpoint calls pass before PR is opened; any failure aborts
• Branch is created off main (no direct push to origin/main)

🤖 Generated with Claude Code

[VinciGit00]

Endpoint validation — extract (5/5 passed)

Run via tests/python-v2.1.0/test_extract.py against scrapegraph-py>=2.1.0. API key supplied by the user at PR time; not stored, logged, or echoed. Response payloads omitted — key names only.

[extract 1/5] https://example.com          -> status=success elapsed_ms=534  keys=['page_title', 'main_heading']
[extract 2/5] https://scrapegraphai.com    -> status=success elapsed_ms=1529 keys=['answer', 'description']
[extract 3/5] https://www.iana.org/        -> status=success elapsed_ms=1234 keys=['main_purpose']
[extract 4/5] https://example.org          -> status=success elapsed_ms=348  keys=['title', 'description']
[extract 5/5] https://httpbin.org/html     -> status=success elapsed_ms=1421 keys=['summary']

Demonstrates the new step 4 added in 099594e.